Концепция
Регресионният анализ е математически модел. Когато зависимата променлива и независимата променлива имат линейна връзка, това е специален линеен модел.
Най-простият случай е линейна регресия с една променлива, която се състои от независима променлива и зависима променлива, които са приблизително линейно свързани; моделът е Y=a+bX+ε(X е независимата променлива, Y е променливата-причина, ε е случайна грешка).
Usuallyassumethatthemeanvalueoftherandomerroris0,andthevarianceisσ^2(σ^2﹥0,σ^2hasnothingtodowiththevalueofX).Ifitisfurtherassumedthattherandomerrorfollowsanormaldistribution,itiscalledanormallinearmodel.Generally,iftherearekindependentvariablesand1dependentvariable,thevalueofthedependentvariableisdividedintotwoparts:onepartisaffectedbytheindependentvariable,thatis,expressedasitsfunction,thefunctionformisknownandcontainsunknownparameters;theotherpartisdeterminedbyOtherunconsideredfactorsandrandomeffectsarerandomerrors.
Whenthefunctionisalinearfunctionwithunknownparameters,itiscalledalinearregressionanalysismodel;whenthefunctionisanonlinearfunctionwithunknownparameters,itiscalledanonlinearregressionanalysismodel.Whenthenumberofindependentvariablesisgreaterthan1,itiscalledmultipleregression,andwhenthenumberofdependentvariablesisgreaterthan1,itiscalledmultipleregression.
Съдържание на регресионния анализ
Основното съдържание на регресионния анализ е следното:
①Startingfromasetofdata,determinethequantitativerelationshipbetweencertainvariables;Thatis,amathematicalmodelisestablishedandunknownparametersareestimated.Usuallytheleastsquaremethodisused.
②Тествайте надеждността на тези отношения.
③Intherelationshipbetweenmultipleindependentvariablesaffectingadependentvariable,judgewhethertheindependentvariablehasasignificantimpact,andselectthesignificantimpactintothemodel,andeliminateinsignificantvariables.Stepwiseregression,forwardregression,andbackwardregressionareusuallyused.
④Usetherequiredrelationshiptopredictorcontrolacertainprocess.
Theapplicationofregressionanalysisisveryextensive,andtheuseofstatisticalsoftwarepackagescanmakevariousalgorithmsmoreconvenient.
Видове регресия
Основните типове регресия са: линейна регресия, криволинейна регресия, двоична логистична регресия и множествена логистична регресия.
Applicationofanalysis
Correlationanalysisstudiesthecorrelationbetweenphenomena,thedirectionandclosenessofcorrelation,andgenerallydoesnotdistinguishbetweenindependentvariablesordependentvariables.Regressionanalysisistoanalyzethespecificformsofcorrelationbetweenphenomena,determinethecausalrelationship,andusemathematicalmodelstoexpressthespecificrelationship.Forexample,fromthecorrelationanalysis,wecanknowthatthe"quality"and"usersatisfaction"variablesarecloselyrelated,butwhichvariablebetweenthesetwovariablesisaffectedbywhichvariable,andthedegreeofinfluence,requiresregressionanalysisMethodtodetermine.
Generallyspeaking,regressionanalysisistodeterminethecausalrelationshipbetweendependentvariablesandindependentvariables,establisharegressionmodel,andsolvethevariousparametersofthemodelbasedonthemeasureddata,andthenevaluatewhethertheregressionmodelisItcanfitthemeasureddatawell;ifitcanfitwell,furtherpredictionscanbemadebasedontheindependentvariables.
Forexample,ifyouwanttostudythecausalrelationshipbetweenqualityandusersatisfaction,inapracticalsense,productqualitywillaffectusersatisfaction,sosetusersatisfactionasthedependentvariableandrecorditasY;Qualityistheindependentvariable,denotedasX.AccordingtothescatterplotinFigure8-3,thefollowinglinearrelationshipcanbeestablished:
Y=A+BX+§
where:AandBareundeterminedparameters,andAisregressionTheinterceptofthestraightline;Bistheslopeoftheregressionline,whichrepresentstheaveragechangeofYwhenXchangesbyoneunit;§istherandomerrortermthatdependsonusersatisfaction.
LinearregressioncanbeeasilyimplementedintheSPSSsoftware.Theregressionequationisasfollows:
;Theexampleshownaboveisasimplelinearregressionproblemofoneindependentvariable.Duringdataanalysis,thiscanalsobeextendedtomultipleregressionofmultipleindependentvariables.PleaserefertothespecificregressionprocessandmeaningRefertorelevantstatisticsbooks.Inaddition,intheSPSSresultoutput,R2,FtestvalueandTtestvaluecanalsobereported.R2isalsocalledthecoefficientofdeterminationoftheequation,whichindicatesthedegreeofinterpretationofthevariableXtoYintheequation.ThevalueofR2isbetween0and1.Thecloserto1,thestrongertheinterpretationabilityofXtoYintheequation.R2isusuallymultipliedby100%toexpressthepercentageofYchangeexplainedbytheregressionequation.TheFtestisoutputthroughtheanalysisofvariancetable,andthesignificancelevelisusedtotestwhetherthelinearrelationshipoftheregressionequationissignificant.Generallyspeaking,significancelevelsbelow0.05aremeaningful.WhentheFtestpasses,itmeansthatatleastoneoftheregressioncoefficientsintheequationissignificant,butnotallregressioncoefficientsaresignificant,soaTtestisneededtoverifythesignificanceoftheregressioncoefficients.Similarly,theTtestcanbedeterminedbythesignificanceleveloralook-uptable.Intheexampleshownabove,themeaningofeachparameterisshowninTable1-1.
Таблица 1-1 тест за линейно регресионно уравнение
индекс | Стойност | Ниво на значимост | Значение |
R | 0,89 | "Качество" обяснява 89% от степента на промяна в "Удовлетвореност на потребителите" | |
F | 276,82 | 0,001 | Линейната връзка на регресионното уравнение е значима |
Т | 16,64 | 0,001 | Коефициентът на регресионното уравнение е значим |